(74 days)
Quantra™ is a software application intended for use with images acquired using digital breast x-ray systems. Quantra calculates volumetric breast density as a ratio of fibroglandular tissue and total breast volume estimates. Quantra also provides area breast density as a ratio of fibroglandular tissue area estimates. Quantra segregates breast density into categories, which may be useful in the reporting of consistent BI-RADS® breast composition categories as mandated by certain state regulations. The Quantra results for each image, breast, and subject, are intended to aid radiologists in the assessment of breast tissue composition. Quantra produces adjunctive information; it is not an interpretive or diagnostic aid.
The Quantra 2.1 software provides volumetric and area assessment on digital x-ray images of the breast. The Quantra software calculates estimates of breast tissue volumes, estimates of breast tissue areas and statistical measures. The Quantra software operates on a Windows server (e.g., Hologic Cenova server) that meets Quantra data input and output requirements and generally is located outside the patient environment. The device does not contact the patient, nor does it control any life-sustaining devices.
Here's an analysis of the acceptance criteria and study information for the Quantra 2.1 device, based on the provided text:
1. Table of Acceptance Criteria and Reported Device Performance
The document does not explicitly state acceptance criteria in the form of numerical thresholds for device performance. Instead, the "Performance/Bench Testing" section describes the types of evaluations performed and the comparisons made. The conclusion states that the information and data presented were "adequate...to determine substantial equivalence to the predicate device." This implies the device's performance was considered acceptable if it was comparable to or improved upon the predicate.
| Acceptance Criteria (Implied) | Reported Device Performance |
|---|---|
| Accuracy of fibroglandular tissue thickness measurements | Assessed using a breast phantom of known composition. (No specific numerical accuracy reported, but implies the measurements were deemed satisfactory for substantial equivalence). |
| Reproducibility of breast density measurement (Vbd) | Assessed using an image dataset of the same patients imaged within a short period of time on two different acquisition systems. (No specific reproducibility metric reported, but implies acceptable consistency for substantial equivalence). |
| Consistency of density measures across different views and breasts | Evaluated statistically between CC and MLO views of the same breast and left and right breasts of the same women, using a large database of images from Hologic (Selenia and Selenia Dimensions), GE (Senographe and Senographe Essential), and Siemens (Mammomat Novation) digital breast x-ray systems. (Implies statistical consistency was found, supporting substantial equivalence). |
| Correlation with expert BI-RADS® category scores | Images in a database including BI-RADS category scores assigned by 15 readers were used to compare the reader BI-RADS scores to the q abd, O abd. Vbd and Abd measures. (Implies a correlation or agreement was sufficient to establish substantial equivalence; no specific correlation coefficient or agreement metric is provided). |
| Equivalence to Predicate Device (Quantra 2.0) | Quantra 2.1 provides the same breast density and statistical assessments as Quantra 2.0. The algorithm is updated for independence from certain acquisition parameters. Includes a new reference population with Hologic Selenia Dimensions images. The method to combine image level results for per-breast and per-patient breast density measures is updated (averaging vs. maximum in Quantra 2.0). Overall deemed substantially equivalent. |
2. Sample Size Used for the Test Set and Data Provenance
- Test Set Sample Size:
- For the accuracy of fibroglandular tissue thickness, a "breast phantom" was used. This would be a single or a small set of phantoms, not a large patient sample size.
- For reproducibility of Vbd, an "image dataset of the same patients" was used. No specific number of patients is provided, but it implies multiple patients.
- For consistency of density measures across views/breasts, a "large database of images" was used. No specific number is given, but it is explicitly stated to be "large."
- For correlation with BI-RADS® scores, a database of "images including BI-RADS category scores" was used. No specific number of images or patients is detailed.
- Data Provenance: The document explicitly mentions images from "Hologic (Selenia and Selenia Dimensions), GE (Senographe and Senographe Essential), and Siemens (Mammomat Novation) digital breast x-ray systems." This indicates data from various manufacturers, suggesting it's likely multi-centric. The origin countries are not specified, but these are internationally recognized manufacturers. The studies appear to be retrospective as they use "image data sets" and "a large database of images."
3. Number of Experts Used to Establish Ground Truth for the Test Set and Their Qualifications
- For the study comparing reader BI-RADS® scores to Quantra measures, 15 readers assigned BI-RADS® category scores.
- Qualifications of Experts: The term "readers" is used in the context of assigning BI-RADS® scores, strongly implying these are radiologists. Their specific years of experience or sub-specialization are not provided in the document.
4. Adjudication Method for the Test Set
- The document states that "BI-RADS category scores assigned by 15 readers were used to compare the reader BI-RADS scores to the q abd, O abd. Vbd and Abd measures." It does not specify an adjudication method like 2+1 or 3+1 for establishing a single consensus ground truth from these 15 readers. It appears the individual reader scores were used directly for comparison against the Quantra output, rather than an adjudicated consensus being formed.
5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study Was Done
- The document describes comparing the software's performance with BI-RADS® scores from "15 readers." While this involves multiple readers and multiple cases, the study described is a comparison of the device's output against human ratings, not a "comparative effectiveness study" in the form of 'readers with AI vs. readers without AI assistance'.
- Effect Size: No effect size of human readers improving with AI vs. without AI assistance is reported. The device is explicitly stated to produce "adjunctive information; it is not an interpretive or diagnostic aid." This reinforces that it's not intended for human-in-the-loop performance measurement for diagnostic improvement.
6. If a Standalone (Algorithm Only Without Human-in-the-Loop Performance) Was Done
- Yes, standalone performance was assessed. The entire "Performance/Bench Testing" section describes evaluations of the Quantra 2.1 algorithm's output independently.
- Measurement accuracy on phantoms.
- Reproducibility on patient datasets.
- Statistical evaluation of density measures across views/breasts.
- Comparison of Quantra's density measures to expert BI-RADS® category scores.
The goal was to demonstrate substantial equivalence, focusing on the algorithm's direct output.
7. The Type of Ground Truth Used
- Mixed types of ground truth were used:
- Known Composition (Phantom): For the measurement accuracy of fibroglandular tissue thickness.
- Expert Consensus/Categorization (Implied): For comparing Quantra's density measures to BI-RADS® scores, the BI-RADS® scores assigned by the 15 readers serve as an expert-driven "ground truth" for categorization.
- Clinical Data (Implied): For reproducibility and consistency evaluations, the actual patient images and their characteristics formed the basis for assessment.
8. The Sample Size for the Training Set
- The document states, "Quantra 2.1 includes a new reference population with the more recent Hologic Selenia Dimensions image sets." However, it does not specify the sample size for this new reference population, nor explicitly differentiate it as a "training set." While a "training set" is generally used for machine learning models, the Quantra device performs calculations based on an algorithm, so the term "reference population" might refer to data used for algorithm development, calibration, or internal validation rather than a traditional machine learning training set.
9. How the Ground Truth for the Training Set Was Established
- The document does not explicitly describe how ground truth was established for any "training set" or "reference population." It describes the use of the reference population (Hologic Selenia Dimensions image sets) but not the process for labeling or establishing ground truth within that set.
§ 892.2050 Medical image management and processing system.
(a)
Identification. A medical image management and processing system is a device that provides one or more capabilities relating to the review and digital processing of medical images for the purposes of interpretation by a trained practitioner of disease detection, diagnosis, or patient management. The software components may provide advanced or complex image processing functions for image manipulation, enhancement, or quantification that are intended for use in the interpretation and analysis of medical images. Advanced image manipulation functions may include image segmentation, multimodality image registration, or 3D visualization. Complex quantitative functions may include semi-automated measurements or time-series measurements.(b)
Classification. Class II (special controls; voluntary standards—Digital Imaging and Communications in Medicine (DICOM) Std., Joint Photographic Experts Group (JPEG) Std., Society of Motion Picture and Television Engineers (SMPTE) Test Pattern).